Introduction
The maternal mortality ratio (MMR) is an important indicator of health care quality in a given society. Despite a decrease in the worldwide MMR in the past 2 decades progress has been much slower than expected. An estimated 303,000 women died of causes related to pregnancy and childbirth in the world in 2015, and the majority (∼99%) of these deaths occurred in developing countries. Yet, in developed countries such as the United States, the MMR has not decreased, with recent studies suggesting an increase in such deaths. Given these factors, the United Nations proposed the reduction of the global MMR to <70 deaths per 100,000 live births by 2030 as a sustainable development goal.
Analyses of maternal deaths have consistently revealed that delays in the recognition of pregnancy complications are associated with higher mortality. Therefore, the use of early warning systems (EWS) have been proposed as a potential tool to reduce maternal morbidity and mortality, based on identification of predetermined abnormal values (warning signs) to generate a rapid and effective medical response. Several EWS have been developed for obstetrical patients, but the majority are the result of a clinical consensus rather than statistical analyses of clinical outcome measures (ie, maternal deaths). In 2013, Carle et al, with the Intensive Care National Audit and Research Center (ICNARC) Case Mix Program in the United Kingdom, designed and internally validated a statistically derived, clinically modified, Obstetric Early Warning Score (OEWS) using 4440 obstetric admission to the intensive care unit (ICU) that showed a good predictive ability to discriminate survivors from nonsurvivors.
Recent studies from the United States have shown that the implementation of a clinical, pathway-specific maternal EWS significantly reduces severe maternal morbidity. There have been no studies assessing the utility of EWS in critically ill obstetric patients for the prediction of maternal death. As maternal death often occurs in the ICU, early warning tools may help allocate appropriate resources in the ICU. The ICNARC-OEWS has been the only obstetric early warning tool designed with a large sample size and a statistically derived score using maternal death as the main outcome. An important goal for the generalizability of any score is the validation in different settings such as populations outside the original development data set. Thus, the objective of this study was to validate the performance of the ICNARC-OEWS for the prediction of maternal death in peripartum women admitted to a critical care unit in a developing country.
Materials and Methods
Study design and setting
A single-center retrospective cohort study was designed with all consecutive pregnant and postpartum (up to 42 days) women who required admission to the ICU Gestion Salud at Clínica de Maternidad Rafael Calvo (CMRC) in Cartagena, Colombia, from Jan. 1, 2006, through Dec. 31, 2011. CMRC is a large, university-affiliated maternity hospital that delivers nearly 9000 women per year, accounting for 46% of all live births in Cartagena, a city of approximately 1 million people. In 2010 the estimated MMR in Colombia was 92 deaths per 100,000 live births. All consecutive pregnant patients and women admitted to the ICU within 42 days postpartum were eligible for the study. The institutional review board of Gestion Salud at CMRC approved the study. All women provided written informed consent for the use of their clinical data for research purposes.
Data collection
Clinical information was gathered from hospital medical records by a trained research nurse, and described elsewhere. Data included demographic variables (maternal age, socioeconomic status), obstetric variables (parity, pregnancy status, and gestational age at time of admission), diagnosis upon admission and during the stay in the ICU, lowest and highest values of vital signs in the first 24 hours after admission (temperature, systolic blood pressure, diastolic blood pressure, heart rate, and respiratory rate), and need for critical care interventions such as mechanical ventilatory support or use of vasoactive drugs during the ICU stay. The Glasgow Coma Scale (GCS) score was calculated at the time of admission to the ICU. If GCS score was unavailable on admission because of sedation, the last GCS score documented before sedation was used.
Definitions
Women were followed up to hospital discharge and maternal death was defined as death during pregnancy or within 42 days postpartum. Direct obstetric-related disorders were defined as those resulting from obstetric complications of the pregnant state (pregnancy, delivery, and postpartum) or chain of events resulting from pregnancy-related disorders (eg, hypertensive disorders in pregnancy). Indirect obstetric causes were defined as those resulting from preexisting conditions or conditions that developed during pregnancy and may have been aggravated by physiologic effects of pregnancy, but were not due to direct obstetric causes.
Obstetric early warning score
The original publication by Carle et al reported the design of a statistically based early warning score (statistical OEWS) and a modification established by the authors based on the weight of each variable, clinically (clinical OEWS). In this study, we analyzed the score recommended by the authors, the clinical OEWS. The OEWS is calculated based on values of the following variables: systolic blood pressure, diastolic blood pressure, respiratory rate, heart rate, fraction of inspired oxygen (FiO 2 ) required to maintain an oxygen saturation ≥96%, temperature, and the level of consciousness ( Supplemental Table 1 ). Specific scores were defined as follows: 0 for routine care, 1-3 in the aggregate score for low-grade response, 4-5 in the aggregate score or 3 in 1 abnormal vital sign for a medium response, and ≥6 in the aggregate score for a high response.
Statistical analysis
All categorical variables were expressed in percentages, and continuous variables in median with interquartile ranges (IQR) as dispersion measure. Comparisons of categorical variables were performed with χ 2 or the Fisher exact test, when appropriate. Continuous variables were compared through analysis of variance or Kruskal-Wallis, depending on whether variables were normally distributed or not. Normality was assessed with the Shapiro-Wilk test. Odds ratios (OR) were reported with 95% confidence interval (CI) for multivariable logistic regression. We constructed the multivariable logistic regression model following the recommendations of Kleinbaum and Klein to specify the multivariable model.
To assess the diagnostic performance of the clinical OEWS, model discrimination was estimated using the area under the receiver operator characteristic curve (AUC), and reported with 95% CI. Although EWS were not designed to predict mortality, the present study assesses the probability of death predicted by the OEWS to evaluate the possible impact of a trigger and how the different classifications proposed by the ICNARC group relate to mortality in our population. A P value <.05 was considered significant for all analyses. Cases with missing data in any variable related to the OEWS were not included in the analyses. All statistical analyses were performed using statistical software (Stata 13; StataCorp, College Station, TX).
Materials and Methods
Study design and setting
A single-center retrospective cohort study was designed with all consecutive pregnant and postpartum (up to 42 days) women who required admission to the ICU Gestion Salud at Clínica de Maternidad Rafael Calvo (CMRC) in Cartagena, Colombia, from Jan. 1, 2006, through Dec. 31, 2011. CMRC is a large, university-affiliated maternity hospital that delivers nearly 9000 women per year, accounting for 46% of all live births in Cartagena, a city of approximately 1 million people. In 2010 the estimated MMR in Colombia was 92 deaths per 100,000 live births. All consecutive pregnant patients and women admitted to the ICU within 42 days postpartum were eligible for the study. The institutional review board of Gestion Salud at CMRC approved the study. All women provided written informed consent for the use of their clinical data for research purposes.
Data collection
Clinical information was gathered from hospital medical records by a trained research nurse, and described elsewhere. Data included demographic variables (maternal age, socioeconomic status), obstetric variables (parity, pregnancy status, and gestational age at time of admission), diagnosis upon admission and during the stay in the ICU, lowest and highest values of vital signs in the first 24 hours after admission (temperature, systolic blood pressure, diastolic blood pressure, heart rate, and respiratory rate), and need for critical care interventions such as mechanical ventilatory support or use of vasoactive drugs during the ICU stay. The Glasgow Coma Scale (GCS) score was calculated at the time of admission to the ICU. If GCS score was unavailable on admission because of sedation, the last GCS score documented before sedation was used.
Definitions
Women were followed up to hospital discharge and maternal death was defined as death during pregnancy or within 42 days postpartum. Direct obstetric-related disorders were defined as those resulting from obstetric complications of the pregnant state (pregnancy, delivery, and postpartum) or chain of events resulting from pregnancy-related disorders (eg, hypertensive disorders in pregnancy). Indirect obstetric causes were defined as those resulting from preexisting conditions or conditions that developed during pregnancy and may have been aggravated by physiologic effects of pregnancy, but were not due to direct obstetric causes.
Obstetric early warning score
The original publication by Carle et al reported the design of a statistically based early warning score (statistical OEWS) and a modification established by the authors based on the weight of each variable, clinically (clinical OEWS). In this study, we analyzed the score recommended by the authors, the clinical OEWS. The OEWS is calculated based on values of the following variables: systolic blood pressure, diastolic blood pressure, respiratory rate, heart rate, fraction of inspired oxygen (FiO 2 ) required to maintain an oxygen saturation ≥96%, temperature, and the level of consciousness ( Supplemental Table 1 ). Specific scores were defined as follows: 0 for routine care, 1-3 in the aggregate score for low-grade response, 4-5 in the aggregate score or 3 in 1 abnormal vital sign for a medium response, and ≥6 in the aggregate score for a high response.
Statistical analysis
All categorical variables were expressed in percentages, and continuous variables in median with interquartile ranges (IQR) as dispersion measure. Comparisons of categorical variables were performed with χ 2 or the Fisher exact test, when appropriate. Continuous variables were compared through analysis of variance or Kruskal-Wallis, depending on whether variables were normally distributed or not. Normality was assessed with the Shapiro-Wilk test. Odds ratios (OR) were reported with 95% confidence interval (CI) for multivariable logistic regression. We constructed the multivariable logistic regression model following the recommendations of Kleinbaum and Klein to specify the multivariable model.
To assess the diagnostic performance of the clinical OEWS, model discrimination was estimated using the area under the receiver operator characteristic curve (AUC), and reported with 95% CI. Although EWS were not designed to predict mortality, the present study assesses the probability of death predicted by the OEWS to evaluate the possible impact of a trigger and how the different classifications proposed by the ICNARC group relate to mortality in our population. A P value <.05 was considered significant for all analyses. Cases with missing data in any variable related to the OEWS were not included in the analyses. All statistical analyses were performed using statistical software (Stata 13; StataCorp, College Station, TX).
Results
Patients’ characteristics
During the study period, 50,897 births were recorded. A total of 724 obstetric patients required admission to critical care, for an ICU admission rate of 14.22 per 1000 deliveries. Incomplete vital sign data were found in 22 subjects (3.0%) of the cohort. Of these 22 women, 2 died (9.1%). However, there were no significant differences in mortality rates between women excluded from the sample due to missing data, and women with complete data. There were 29 deaths in the 702 women with complete data, for a case fatality rate of 4.1% and a MMR of 56.98 deaths per 100,000 live births. Clinical characteristics and vital signs in the first 24 hours of ICU admission according to survival status are described in Table 1 . There were no significant differences in maternal age, nulliparity, and length of ICU stay between women who survived to ICU discharge and those who died ( P > .05). Our data demonstrated that those patients who did not survive were more likely to have hypotension, tachycardia, and a low GCS score compared to women who did survive ( Table 1 ).
| Survivors, n = 673 | Nonsurvivors, n = 29 | P value | |
|---|---|---|---|
| Obstetrical/clinical variables | |||
| Maternal age, years, median (IQR) | 23 (19–29) | 24 (21–27) | .363 |
| Nulliparity, n (%) | 449 (66.7) | 24 (82.8) | .071 |
| Length of ICU stay, days, median (IQR) | 4 (2–5) | 4 (1–19) | .404 |
| Vital signs at admission | |||
| Systolic blood pressure, mm Hg | |||
| Maximum value, median (IQR) | 137 (120–159) | 131 (118–163) | .543 |
| Minimum value, median (IQR) | 113 (98–128) | 84 (73–98) | <.001 |
| Diastolic blood pressure, mm Hg | |||
| Maximum value, median (IQR) | 96 (81–110) | 89 (72–101) | .084 |
| Minimum value, median (IQR) | 73 (60–86) | 49 (37–70) | <.001 |
| Respiratory rate/min | |||
| Maximum value, median (IQR) | 22 (19–25) | 22 (20–25) | .290 |
| Minimum value, median (IQR) | 16 (14–18) | 16 (15–18) | .297 |
| Heart rate/min | |||
| Maximum value, median (IQR) | 103 (89–120) | 118 (106–132) | .003 |
| Minimum value, median (IQR) | 78 (68–91) | 85 (68–112) | .222 |
| Temperature, °C | |||
| Maximum value, median (IQR) | 37 (36.7–37) | 36.8 (36–37) | .017 |
| Minimum value, median (IQR) | 36.4 (36–36.6) | 36 (35–36.5) | <.001 |
| Glasgow Coma Scale score, median (IQR) | 15 (15–15) | 14 (6–15) | <.001 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
